Looks to me like the docs are wrong.
Here's from playing around with smoothstep:
y = smoothstep(1.0,-1.0,x);

y = smoothstep(-1.0,1.0,x);

It looks like when edge0 > edge1, it flips the side at 1 to be at negative infinity, and the side at 0 to be at positive infinity.
Another example:
#ifdef GL_ES
precision mediump float;
#endif
#define PI 3.14159265359
uniform vec2 u_resolution;
float plot(vec2 st, float pct){
return smoothstep( pct+0.02, pct, st.y) -
smoothstep( pct, pct-0.02, st.y);
}
void main() {
vec2 st = gl_FragCoord.xy/u_resolution;
// Smooth interpolation between 0.1 and 0.9
float y = smoothstep(0.1,0.9,st.x);
vec3 color = vec3(y);
float pct = plot(st,y);
color = (1.0-pct)*color+pct*vec3(0.0,1.0,0.0);
gl_FragColor = vec4(color,1.0);
}

Changing y to a step from 0.9 to 0.1 changes the output to this:
